Example code for http://vfxcode.blogspot.co.nz/ cuda intro

cudaTest.cpp

#include "cudaTest.h"

CudaTest::CudaTest()
{
}

void* CudaTest::creator()
{
    return new CudaTest;
}

void incrementArrayOnHost(float *a, int N)
{
    int i;
    for (i=0; i<N; i++)
    {
        a[i] = a[i]+1.f;
    }
}


extern "C"
float * test(float *a, int N);



MStatus CudaTest::doIt( const MArgList& argList )
{
    MGlobal::displayInfo( "Testing" );

    float *a_h, *b_h; // host memory pointers
    int N = 100;       // number of arrays
    int i;

    // allocate arrays on host
    a_h = (float *)malloc(sizeof(float)*N);
    b_h = (float *)malloc(sizeof(float)*N);

    // initialize host data
    for (i=0; i<N; i++){
        a_h[i] = (float)i;
    }

    // Calculate on Host
    incrementArrayOnHost(a_h, N);

    // Calculate on Device
    b_h = test(a_h, N);

    // Check Result
    for (i=0; i<N; i++)
        assert(a_h[i] == b_h[i]);

    //cleanup
    free(a_h);
    free(b_h);
    MGlobal::displayInfo( "Done" );

    return MS::kSuccess;
}

cudaTest.cu

// Includes, system
#include <stdio.h>
#include <assert.h>

// CUDA runtime
#include <cuda_runtime.h>


__global__ void incrementArrayOnDevice(float *a, int N)
{
    int idx = blockIdx.x*blockDim.x + threadIdx.x;
    if (idx<N)
    {
        a[idx] = a[idx]+1.f;
    }
}


extern "C"
float * test(float *a_h, int N)
{
    float *b_h; // host memory pointers
    float *a_d; // device memory pointers

    // allocate arrays on device
    cudaMalloc((void **) &a_d, sizeof(float)*N);
    b_h = (float *)malloc(sizeof(float)*N);

    // send data from host to device: a_h to a_d
    // cudaMemCpy(target, source, memToAllocate, function)
    cudaMemcpy(a_d, a_h, sizeof(float)*N, cudaMemcpyHostToDevice);

    int blockSize = 4;
    int nBlocks = N/blockSize + (N%blockSize == 0?0:1);

    incrementArrayOnDevice <<< nBlocks, blockSize >>> (a_d, N);

    // copy data from device to host
    cudaMemcpy(b_h, a_d, sizeof(float)*N, cudaMemcpyDeviceToHost);

    cudaFree(a_d);

    return b_h;
}

cudaTest.h

#ifndef CUDATEST_H
#define CUDATEST_H

// APP
#include <stdio.h>
#include <string.h>
#include <assert.h>

#include <maya/MPxCommand.h>
#include <maya/MGlobal.h>
#include <maya/MObject.h>


class CudaTest : public MPxCommand
{
public:
    CudaTest();
    virtual MStatus doIt( const MArgList& argList );
    static void* creator();
};

#endif

pluginMain.cpp

#include "cudaTest.h"

#include <maya/MFnPlugin.h>

MStatus initializePlugin( MObject obj )
{
    MStatus status;

    MFnPlugin fnPlugin( obj, "Alex Telford", "1.0", "Any" );

    status = fnPlugin.registerCommand( "cudaTest", CudaTest::creator );
    CHECK_MSTATUS_AND_RETURN_IT( status );

    return MS::kSuccess;
}


MStatus uninitializePlugin( MObject obj )
{
    MStatus status;

    MFnPlugin fnPlugin( obj );

    status = fnPlugin.deregisterCommand( "cudaTest" );
    CHECK_MSTATUS_AND_RETURN_IT( status );

    return MS::kSuccess;
}